Transit isochrones incorrect (very underestimated) just South of Paris

[kofronpi]

Hello,

Requests for transit isochrones south of Paris (tested in Saclay, Guyancourt...) give me wrong results ( small isochrones, very underestimated accessibility: Paris never reached although it can be)

Here is a sample request:

https://api.navitia.io/v1/coverage/fr-idf/isochrones?boundary_duration%5B%5D=2700&datetime=20180522T080000&forbidden_uris%5B%5D=physical_mode%3AAir&forbidden_uris%5B%5D=physical_mode%3ATaxi&from=2.072035%3B48.769893

And the isochrone as a result: (40minutes)

And for 60 minutes:

It's the same either with "from" or "to" parameters (departure after / arrival before), and with other dates (I first thought it was because next Monday was a public holiday).

Thanks for your help.

[pbougue]

Hello,

I had a brief look at it, but I do not manage to go at 8am from your starting point to a point in Paris before 9am, even when computing a journey (walking as fallback mode, and using public-transport).

What type of journey do you have in mind? A journey from that exact coordinates arriving before 9am (8am + 3600s), to where in Paris and using what transport?

I see some reasons why it wouldn't work correctly (problem in transport data, in street-network data, bug), but I don't see how to investigate without a simple journey example that you think should work.
You can have a look at that section of doc (if it's not already the case) : https://doc.navitia.io/#isochrones-currently-in-beta especially the part "Understand the resulting isochrone"

Please also keep in mind that starting from specific coordinates, Navitia first computes a walking fallback (one have to walk to the station, and it can be quite long). If you provide a "city id" as a starting point, result could be wider because (most of the time) walking is not considered necessary in that case : https://api.navitia.io/v1/coverage/fr-idf/isochrones?boundary_duration%5B%5D=3600&datetime=20180522T080000&forbidden_uris%5B%5D=physical_mode%3AAir&forbidden_uris%5B%5D=physical_mode%3ATaxi&from=admin%3Afr%3A78297&

[kofronpi]

@pbougue an example from navitia's journey API:

Paris 15 -> Versailles

1. 7:17 am train C - Gare de Versailles Château - Rive Gauche (Versailles) from Gare d'Issy Val de Seine (Issy-les-Moulineaux) at 07:17 to Gare de Versailles Château - Rive Gauche (Versailles) at 07:35 during 18min
2. BUS 39-12 - technocentre Trou Berger (Guyancourt) from Gare de Versailles-Chantiers (Versailles) at 07:35 to Frères Perret (Guyancourt) at 07:52 during 17min

or Versailles -> Paris 15

1. Bus 440 - Versailles Château - Rive Gauche Gare (Versailles) from Dampierre (Guyancourt) at 07:20 to Versailles Château - Rive Gauche Gare (Versailles) at 07:33 during 13min
2. Train C - Gare de Versailles Chantiers (Versailles) from Gare de Versailles Château - Rive Gauche (Versailles) at 07:55 to Gare d'Issy (Issy-les-Moulineaux) at 08:10 during 15min

Even with walking times included, and in both directions, Paris 15 should be roughly accessible and it's far from there. I took the farthest city from the starting point to show that other, closer cities should be included too.

Also, lately I noted a tendency to regularly "skip" cities accessible by TER in isochrones results.
Thank you for your help and I'm willing to help dig more into it :)

[pbougue]

@kofronpi if you have the exact Navitia call for journeys (or any call), it would be easier to talk/debug.
Anyway, from what you tell here is what I can explain.

First (it's in doc maybe not clear enough, we have to see if it can be improved) :

1. When you ask for an 1-hour isochrone starting at 8am, Navitia stops at 9am sharp. The actual time it took is not considered, especially if you go pick your train in 30min because it doesn't go before that. It maybe takes 35min to go there, but as you wait 30min before going, you are too late to be in the 1-hour isochrone.
2. I you start from a precise point, you have to consider the time to get to the station walking. Also, Navitia considers that you can't walk more than 30min to go to a station.

So, from what you say:

• The 2. might explain why you don't get Paris 15->Versailles in isochrone... Issy is too far away from Paris (when walking from a city, we use the city hall as starting point).
  So I get a small part of Guyancourt on a 7am isochrone from "Paris" when allowing arriving until 8:10am : https://api.navitia.io/v1/coverage/fr-idf/isochrones?from=admin%3Afr%3A75056&datetime=20180604T070000&boundary_duration%5B%5D=3600&boundary_duration%5B%5D=3900&boundary_duration%5B%5D=4200&
• The 1. might explain Versailles->Paris : you arrive after 8am which is the actual hard constraint for isochrone.
  But when starting from Dampierre stop at 7am we reach Issy before 8am (it works even better if we use "Guyancourt city") : https://api.navitia.io/v1/coverage/fr-idf/isochrones?from=stop_area%3AOIF%3ASA%3A24%3A15204&datetime=20180604T070000&boundary_duration%5B%5D=3600&boundary_duration%5B%5D=3900&boundary_duration%5B%5D=4200&
• As for the TER being skipped, not sure that all the data are provided by IdFMobilités, this might explain the behavior... Again, you can try to do the classic journey call to test if we have data (still the "Understand the resulting isochrone" part of documentation).

[kofronpi]

@pbougue thanks for your help, I will have a look into this.